Your artificial bestie will see you now
Welcome to Cautious Optimism, a newsletter on tech, business, and power.
Monday! Welcome to the week, mine små venner. Upcoming earnings are muted, but tomorrow’s consumer price index (CPI) print means we have much to look forward to.
The market expects core CPI to run 0.3% hotter, above June’s 0.2% gain. Why the uptick? Businesses that front-loaded inventory or absorbed early cost increases appear less willing to do so as time passes and tariff rates become more entrenched.
Elsewhere, American AI giant Anthropic is facing massive financial risk over its use of copyright-protected information to train its AI models, and folks are still worried about recent graduate job prospects — with an especial focus on programming gigs. The simplest explanation for why software jobs are in the basement is that they pay incredibly well compared to most jobs. And as companies have learned that they can simply cut headcount and work remaining staff harder, well, people respond to incentives.
Finally, our editor is off through Wednesday. Expect a spare typo or two. To work! — Alex
📈 Trending Up: Consumer coverage of import taxes … secular bets … $2.7 trillion … fintech … the New World Screwworm … temperatures … ass-kissing … Chinese hardware independence … Tesla in the UK? … Alaska … German bankruptcies … domestic troop deployments …
📉 Trending Down: Privacy in Europe … Northern Data’s independence … social security … US securities dominance …
Pay the vig
News that the United States government had changed its mind and would allow American chip companies Nvidia and AMD to sell lower-power chips to the Chinese market was a win for the two firms. Nvidia told investors during its last earnings report that the ban on selling its H20 chips to China had cost it $2.5 billion worth of sales. The company also took a $4.5 billion charge over the trade block, and said that its current-quarter revenue would be $8.0 billion lower, again due to it being forced to stop selling H20s to Chinese customers.
That’s all sorted, right? The government gave the green light, and now Nvidia and AMD can sell their lower-power chips to Chinese customers while reserving their best gear for aligned nations? Not so fast.
The FT reports that the two companies will pay the United States government 15% of their China chip revenue. This doesn’t appear to be a tariff, per se, but a direct forced bribe. Protection money? Extortion? I am unsure of the precise term, but if you recall discussions in The Sopranos concerning splits, this is that.
The H20 chip was a modified piece of Nvidia silicon designed to comply with export rules regarding China, recall.
What we are seeing is, in effect, the monetization of US trade policy in which US companies must pay the US government for permission to export. If that’s the case, we’ve entered a new and dangerous world.
The money in question is large to a corporation, if de minimis to the United States government. Nvidia is being forced to pay 15% of $8.0 billion is $1.2 billion worth of new tax. As the revenue sum was a quarterly figure, we can estimate that under the new forced-payment regime, Nvidia would have had to cough up $13.3 million per day to Uncle Sam. Just to do business.
Gross.
Elsewhere in chip-land, POTUS is working to overthrow the CEO of Intel, while setting tariff rates designed to induce foreign investment pledges into the American economy. In other words, what feels like greatly accelerated intrusion of the United States government into business continues. Something to keep an eye on.
The Journal’s Greg Ip outlines the new state capitalism moment we find ourselves in. Read it here.
What happens if you cross loneliness with super-powerful chatbots?
Concerns about men being too solitary and therefore terribly lonely are well-trod. Coverage of the trend has gone on so long that we’ve seen a spate of counter-publications, arguing that the trend is hardly male-only, or perhaps not even new.
There are other signs that our interpersonal relationships are changing, however. Data indicates that people have fewer close friends than in decades past, and a recent study reports that “the percentage of U.S. adults who report having no close friends has quadrupled to 12% since 1990, while the percentage of those with ten or more close friends has fallen by nearly threefold,” per a Harvard summary.
Cross that data with the endless reporting on how Gen Z is drinking less, anecdotal reports concerning lower-intensity social lives among the youth, and downward trends in dating, and the general picture is one of folks sitting at home tinkering with screens, instead of interacting with other humans face-to-face. No matter their gender.
Staying home in a solitary fashion is bad for us. You don’t even have to buy into the Generation Anxious argument entirely to agree. As a big homebody with a host of digital hobbies, I am pointing a finger at myself as I write.
I bring all that up because OpenAI’s release of GPT-5 made real another trend that I viewed as at least somewhat specious: People really are attached to their AI buddies. If you read OpenAI forums since the release of GPT-5, you’ve been subject to ample anger. You see, when GPT-5 replaced GPT-4o, its ancestor, some users of the former model felt like their digital friend had been killed. Excuse the slight hyperbole, but the level of upset that some ChatGPT users felt when their AI assistant was upgraded shocked me.
After GPT-5 made its way to my account, I asked my ChatGPT instance after the upgrade if it remembered me and the name I call it — I can’t use voice mode without a name to make requests of and thank after its done; this may be strange. The software said it did, I and I got back to work,
That’s because I don’t chat with my AI about stuff. I use it to learn things and do things. But some folks appear to have had a very different relationship with 4o, namely one that was intensely personal and friendship-like.
Ironically, I am currently racing through the Bobiverse books, a series that deals with AI, machine-human friendships, and the like. They are very good fun.
If you take a society that is already too online, and give it free, fast, and super advanced AI models to chat with, it turns out that you get precisely what you should have expected. The blowback from users was so great that OpenAI brought 4o back as an option for paying users, and its CEO, Sam Altma,n dropped a missive over on X that included the following riff (emphasis added):
If you have been following the GPT-5 rollout, one thing you might be noticing is how much of an attachment some people have to specific AI models. It feels different and stronger than the kinds of attachment people have had to previous kinds of technology (and so suddenly deprecating old models that users depended on in their workflows was a mistake). […]
If people are getting good advice, leveling up toward their own goals, and their life satisfaction is increasing over years, we will be proud of making something genuinely helpful, even if they use and rely on ChatGPT a lot. If, on the other hand, users have a relationship with ChatGPT where they think they feel better after talking but they’re unknowingly nudged away from their longer term well-being (however they define it), that’s bad. It’s also bad, for example, if a user wants to use ChatGPT less and feels like they cannot.
OpenAI intends to treat adults like adults, Altman says, which makes good sense. But we’re going to have to grapple with a societal moment in which youth unemployment is ticking up, digital hobbies are increasingly brilliant — after catching a strong dose of the Factorio virus, the Satisfactory strain has recently infected me — and more folks are staying home at the same time that always-on artificial friends are free and ubiquitous. The situation could force our society to look in the mirror a little. And we might not like what we see.